How do language models use information provided as context when generating a response? Can we infer whether a particular generated statement is actually grounded in the context, a misinterpretation, or fabricated? To help answer these questions, we introduce the problem of context attribution: pinpointing the parts of the context (if any) that led a model to generate a particular statement. We then present ContextCite, a simple and scalable method for context attribution that can be applied on top of any existing language model. Finally, we showcase the utility of ContextCite through three applications: (1) helping verify generated statements (2) improving response quality by pruning the context and (3) detecting poisoning attacks. We provide code for ContextCite at https://github.com/MadryLab/context-cite.
@article{arxiv.2409.00729,
title = {ContextCite: Attributing Model Generation to Context},
author = {Benjamin Cohen-Wang and Harshay Shah and Kristian Georgiev and Aleksander Madry},
journal= {arXiv preprint arXiv:2409.00729},
year = {2024}
}